Cryptojacking is the unauthorized use of someone else’s computing resources to mine cryptocurrency. AI servers can be tempting targets because GPUs and machine-learning instances offer powerful parallel compute—but the cloud incidents documented by Microsoft and AWS point chiefly to compromised credentials and excessive permissions, not to AI workloads being inherently insecure or public exposure alone causing an attack.
What is cryptojacking?
In cloud cryptojacking, an attacker gains or misuses access to a cloud account, creates or takes over computing resources, and runs mining software that contributes work to a cryptocurrency mining pool. The resources belong to a legitimate tenant, so provisioning can resemble ordinary cloud activity until someone notices unusual use, charges, or control-plane events.
| # | Preview | Product | Price | |
|---|---|---|---|---|
| 1 |
|
FortiGate-40F Firewall Appliance plus 1 Year FortiCare Premium and FortiGuard Unified Threat... | Buy on Amazon |
As an Amazon Associate I earn from qualifying purchases.
The immediate harm is resource theft: unexpected bills, capacity diverted from legitimate training or inference, and possible service disruption. A compromised cloud environment can also become a foothold for persistence, lateral movement, or information theft, according to Microsoft Threat Intelligence.
Free tools Windows power users keep installed
One-click scans. No signup required.
Why can AI servers appeal to cryptojackers?
Many AI systems use GPUs or high-performance machine-learning instances. Their parallel compute can be useful for some mining workloads, making that capacity valuable to an attacker who has gained control. Microsoft reported abuse involving Azure T4, V100, and A100 GPU instances; AWS described a campaign targeting GPU and ML instance families. These findings show that such compute can be repurposed, not that AI servers are uniquely targeted or less secure by nature.
#1 Best Overall
- INTEGRATED FIREWALL APPLIANCE AND SECURITY SERVICES: Comes with FortiGate-40F Firewall Appliance, 1 year of FortiCare Premium, and FortiGuard Unified Threat Protection.
- UTP SECURITY FEATURES: Offers protection from advanced threats with DNS filtering, URL filtering, video filtering, and controls against botnets.
- IDEAL FOR SMALLER SETTINGS: Best suited for small to mid-sized businesses needing reliable security without the complexity of larger systems.
- CONTINUOUS SUPPORT AND MAINTENANCE: FortiCare Premium ensures that technical help is readily available to manage and troubleshoot issues.
- COMPACT AND EFFECTIVE: Provides a powerful, yet compact security solution that effectively protects against a wide range of cyber threats.
The distinction matters: “exposed” may mean a public-facing API or management dashboard, while cloud cryptojacking reports also describe attackers using stolen credentials and valid cloud APIs. A public endpoint can enlarge an organization’s attack surface, but the cited incidents do not establish public exposure by itself as the cause of the mining activity.
What the reported figures do—and don’t—show
Microsoft said the incidents it investigated involved more than $300,000 in compute fees. That is an observed total in those cases, not a typical loss estimate. Microsoft also published historical Ethereum Proof of Work mining rates based on February 2023 network complexity: 25.1 MH/s for Azure NC T4 v3 with an NVIDIA T4, 89.5 MH/s for NCv3 with an NVIDIA V100, and 175 MH/s for ND A100 v4 with an NVIDIA A100 40GB. These figures are historical technical context, not a current mining-profitability comparison. See Microsoft’s cloud resource abuse analysis.
How a cloud cryptojacking attack can unfold
- Gain access. An attacker obtains credentials or otherwise abuses access to a cloud tenant. Microsoft described credential compromise and missing multifactor authentication (MFA) in observed incidents.
- Explore permissions and capacity. The attacker checks what the identity can do, what quotas are available, and where resources can be created.
- Provision or take over compute. Existing permissions may let the attacker create resources across regions, use automation, or expand capacity. Microsoft also described use of GPU driver extensions to speed deployment.
- Run mining software. The compute is configured to mine and connect to a pool, consuming capacity that may have been intended for AI work or other services.
A December 2025 AWS report described a campaign against Amazon EC2 and Amazon ECS that began on November 2, 2025. The actor used compromised IAM credentials, enumerated permissions and quotas, and deployed mining resources; AWS said miners were operational within ten minutes of initial access. AWS explicitly reported no AWS service vulnerability in that campaign: the activity used valid credentials without authorization. Read the AWS Security Blog account.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitchesWarning signs to investigate
No single signal proves cryptojacking. Investigate combinations of identity, provisioning, workload, network, and billing evidence rather than relying only on a sudden increase in GPU utilization.
Quick Recap
- Unexpected GPU or ML provisioning: bursts of new instances, particularly from identities that do not normally create compute.
- Quota or regional anomalies: unfamiliar quota checks or increases, capacity exhausted across regions, or autoscaling groups nobody recognizes.
- Suspicious VM extensions: unexpected GPU driver extensions or repeated attempts to install them on unsupported virtual machines. Microsoft Defender for Cloud documents alerts for Azure VM extension activity; coverage depends on the service plan and configuration.
- Mining-related network activity: workloads connecting to mining pools, especially alongside unexplained utilization or cost changes. Microsoft calls pool connections a strong compromise indicator in its described context, but responders should validate surrounding telemetry.
- Unusual identity or API behavior: anomalous administrator sign-ins, unfamiliar locations, unexpected permission checks, or automated cloud API use. AWS describes account and EC2/ECS activity indicators in its campaign report.
How to reduce the risk
Protect identities and limit permissions
- Require strong MFA for privileged cloud accounts, use unique credentials, handle secrets carefully, and remove unused credentials. Microsoft reported that almost all accounts in its observed incidents lacked MFA.
- Apply least privilege: grant only the permissions needed, and restrict who can create, expand, or administer GPU and ML capacity.
Control provisioning, spend, and visibility
- Review compute quotas and alert on unusual GPU/ML provisioning, quota changes, and spending. Cost alerts can surface abuse, but should complement security monitoring rather than replace it.
- Monitor cloud audit and control-plane events, GPU extensions, workload processes, and outbound network behavior. Confirm which provider detections are available under your plan and how they must be configured.
Reduce exposure and verify software sources
- Keep AI services and management interfaces behind appropriate access controls, patch exposed services, and avoid unnecessary internet-facing components. CISA’s joint guidance on deploying AI systems securely frames protection, detection, and response as complementary safeguards for AI systems and related services.
- Be alert to malware delivered through fake software downloads as well as cloud-account abuse. In a May 2026 report, Microsoft described a cryptojacking campaign using fake utility download sites and instances where chatbot interactions were associated with malicious download recommendations. Verify downloads through vendor-controlled sources; see Microsoft’s campaign report.
What to do if you suspect mining
- Follow your cloud provider’s incident procedures and secure or revoke credentials that may be compromised.
- Contain unauthorized compute and review audit logs, resource changes, quota activity, and sign-ins to establish what was created and how access was obtained.
- Check for persistence and lateral movement before treating the incident as resolved. Adapt containment and recovery to your provider, services, and organization’s response plan.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
